DC power versus AC power
blitz
blitz at macronet.net
Mon Dec 30 01:59:41 UTC 2002
Just some musings...
Been watching this discussion for a couple loops now. I have to say you're
both right on certain things, and that each individual design has to be
done on the merits of the need for 100% uptime vs what's tolerable.
AC is always easier to run, as the conductors are smaller because the
voltage is higher and I/R losses are smaller. It does not however, lend
itself to staying alive in that time when the AC mains fail and the
generator starts and gets itself online. For that one needs a battery.
From my days at one of the three letter telcos, we had huge battery banks,
monstrous individual cells, that had ampacities that you can arc weld with
with ease and never notice the drain. However, the conductors running
around the sites were often the size of your wrist and the I/R drop in even
these conductors was noticeable, especially when you had two ESS switches
plus the transport gear on battery alone. We backed up the battery plant(s)
with a 1 megawatt Cat diesel.
Both AC supply and DC have their places, both have limitations.
Obviously if its got to always be there-no-excuses-no-margin-for error, the
costly battery solution is the only way. Batteries, copper, chargers, fuse
panels, big wires. $$$$$$
Add maintenance.. yes, water boy...someone's going to have to check
tightness of all those connections, as even a little looseness will turn a
buss bar into molten metal at the amperages we are talking about here.
Someone will need to test the capacity of the system from time to time
under load conditions as well, to find cells going bad, and hot spots in
the conductors. Liquid cells have specific gravity charts that need watched
and water added as it evaporates. This makes them a maintenance job at
regular intervals.
"If you don't do this now, under good times and daylight conditions, you
WILL be doomed to do it under power failure, and no light conditions". -Sparky
One last thought on DC safety...there is a LOT of current in even one of
these cells. I've seen a finger amputated when a wedding ring bridged a
pair of conductors and a metal watch band vaporize when laid across a buss
bar, resulting in 3rd degree burns. Typical voltages are 48v which is just
low enough to give a sweaty arm a tingle to let you know the trons are
there. Bridge them with a conductor and expect fireworks. Anyone working
around batteries needs to have their head screwed on right. They are not
typically dangerous, but you do have to make sure you observe the rules.
You have high current, dangerous acid, and hydrogen gas present. You should
treat it as you would any potentially dangerous industrial process, with
respect.
Scale all this to the size of the battery plant you're contemplating using
to protect your gear.
If I were designing a site, and had unlimited $, a battery plant would back
it up. We stored 48 hours of electricity, enough time to repair or replace
the generator.
If you can rely on your power feed in being there within a few seconds of a
commercial mains failure, (your site has an adequate generator, which is
maintained and tested regularly) you can usually get by with that lowball
UPS system, one that youre going to lock away in a room or the bottom of a
rack and forget about until the feces hits the air mover. You will, don't
argue, I know you will. When the power's on, you don't pay attention to
backup systems, please see "sparky" quote above.
I've even had idiots say "we cant test the UPS because, What happens if it
fails? we go down"....cant do that in the middle of a business day... yet
they would never allocate some "night" time to test. Ok, see you next outage.
My idea of a UPS is take the maximum amount of power you need for your
equipment now, and add 50% then double it.
The 50% is normal equipment additions, the double is for that weak battery
two or three years down the road.
Ive run an ISP for three days during an Ice storm, using 4 tractor trailer
batteries, and a great, big, over-rated UPS. Two batteries were online at
one time, in parallel, two were being charged from a 150 amp alternator on
my truck to replace them in 4 hours. The secret here was plenty of battery,
plenty of UPS. Never let the battery go dead, and be able to change out the
parallel battery without loosing all power. A simple matter of planning.
That little gel-cell inside your UPS is a 12 volt battery in all aspects. A
tractor trailer battery is as well. You can always use a larger battery as
room permits. If the UPS can handle the load, all you need to feed it is
12v, car battery, truck battery, etc. all work. Just observe polarity, as
you'll only get one chance if you do it wrong.
A word or two about Gell-Cells:
Gell cells are nice, and don't slop like liquid cells. There is a price you
pay. Gelatin electrolytes have what's called an electron-mobility problem
caused by the "Gelatin" in the electrolyte. Less current can be consumed at
once, and inversely, recharge rates have to be lower because of this. You
can't give a gell-cell a good stiff 100 amp recharge when theyre almost
dead, while you CAN do this to a liquid electrolyte battery like's in your car.
So, power protection depends on understanding what youve got to do and what
you can afford. There are, as with most all things, trade off's. Its
entirely possible to adequately protect your entire site, a co-lo, or your
desktop computer given the right equipment. Do it right, and maintain and
test.
Marc
macronet.net
At 14:10 12/29/02 -0600, you wrote:
> > Switched power supplies really don't care about the quality of the
> > sine waves that feeds them, as long as they have energy to put into
> > the "tank". On the other hand, video monitors like sine waves, and
> > they may not get along with DC inverters/rectifiers (or even portable
> > AC no-breaks, which usually generates AC from DC).
>
>Sure. And not to get into a contest here, but it all depends on how bad the
>approximation is and what gear it is we're talking about.
>
>It's just something to consider and research, and include in the decision.
>Clean power plants (and that's got nothing to do with green ;) is always a
>good thing. Doesn't have to be perfect, just close enough for <insert
>operational criteria here>.
>
>Thanks,
>Christian
More information about the NANOG
mailing list